Suitability evaluation of precipitation data using SWAT model
Zhang Xiaoli,Peng Yong,Wang Bende,Wang Haixia
DOI: https://doi.org/10.3969/j.issn.1002-6819.2014.19.011
2014-01-01
Abstract:Precipitation is the important forcing data for hydrological models. However, the precision of hydrological modeling in many regions of the world is limited by the lack of precipitation data. To overcome such limitation, some available globally gridded high resolution precipitation datasets have been used to simulate runoff. Although these precipitation products have been proved to have certain accuracy and good potential for hydrologic simulation, selecting the most optimal precipitation product for certain regions is still important because the accuracy and hydrological utility of precipitation dataset vary in different regions. In this study, the suitability of three precipitation products: GLDAS, TMPA and ERA-Interim for hydrological modeling are evaluated using the multi-objective fuzzy optimum model and the Soil and Water Assessment Tool (SWAT) model during the period of 2001-2012 in the Huanren reservoir catchment, located in Hunjiang River of China. SWAT is a physical-based hydrological model, which takes precipitation series from gauged stations as input to simulate runoff. First, in order to obtain the rainfall time series for SWAT model, three interpolation methods, including inverse distance-weighted method, bilinear interpolation method and the nearest point method, were used to compute the station data from the three precipitation products. Then, the quantitative accuracy of these time series were assessed by statistical indices of mean error (ME), root mean square error (RMSE), correlation coefficient (CC) and relative bias (Bias). The multi-objective optimum fuzzy model aiming to identify the appropriate interpolation method and justify the precipitation products alternative to the gauged values was established, which used ME, RMSE, CC and Bias as characteristic indices. The results revealed that the precipitation of GLDAS cannot serve as an alternative due to its underestimation for the whole basin, which can increase the risk of flood control operation. Moreover, the bilinear interpolation was more appropriate to interpolate the grid precipitation to rain gauge compared with the inverse distance-weighted method and the nearest point method. In addition the precipitation of TMPA and ERA-Interim obtained with the bilinear interpolation method were proved to have perfect fitting with the gauge values at daily and monthly scale, thus they both were alternative to the gauge values as the forcing data to the hydrological model. The two precipitation series were then used as the input to the SWAT model that has been calibrated with rain gauge inputs to get the simulated stream flow and their Bias. Correlation coefficient and Nash-Sutcliffe efficiency coefficient (NSE) were calculated against the observed stream flow. Based on the criteria of the statistical indices specified in the present study, it can be concluded that the performances of these simulations were acceptable at daily and monthly scale. With Bias, CC and NSE as characteristic indices, the multi-objective optimum fuzzy model was developed to evaluation the suitability of TMPA and ERA-Interim for hydrological simulation. The results revealed that TMPA was more suitable for hydrological runoff modelling at the basin scale. The methodologies and approaches developed in the present study will be a reference for hydrological modelling of similar watersheds.